Cerebus Restoration Replacement Signature Recap

Sean Michael Robinson:

The new Cerebus volumes have now made their way into civilian hands. And so, as Dave requested, it's time for a little tour through our work on the book. Mara and I put together the following pages--

9 - 32
97 - 160
289 - 320

This is the only portion of the book that we worked on. The remainder of the book is the work of George Gatsis. You can read my appraisal of that portion of the work here.

It's easiest to do these comparisons from print rather than on a screen. So prop yourself up with your old Cerebus phone book, and let's do a side-by-side.

*********************************************************************************

Let me dispense with the self-congratulation first. The signatures we prepped for this book look as good or better than these pages have ever looked in print.

I'm tempted to say “better” across the board, but that's only really true if you have only read these in collected form. For the pages sourced from the original pamphlets, such as most of issue one, these versions are carbon copies of those earlier printings, except with more consistent black coverage, on better paper, along with some other changes I'll talk about below.

As soon as the file copies of Cerebus 1-25 arrived at the studio, as soon as we'd cracked a cover or two, it was clear that we would be sourcing our scans from these. For whatever reason-- multi-generational negatives, deteriorating printing plates, who knows- it was a night and day difference for several of the issues. The 1987 Cerebus collected volume features significant fill-in of fine detailed hatching, clumpy, dark screen-tone, and areas of blown-out detail that are present and accounted for in the initial printings.

So for the majority of these pages, having no negatives and very little original art, we had to do our best to replicate what was on the page of those initial printings, and improve on them in any areas we could.

Page 10, as it appears in the 1994 Cerebus phone book. Notice the dot gain, the dirty tone, all of the noise in the blacks, and the "corrected" text. Whose touch-up lettering is this? Certainly not Dave's.

 Our page ten, after processing but before any cleanup or correction. This page was sourced directly from one of Dave's file copies, which still has Deni's handwriting in corners of some of the pages. Looking at the lettering, you can see why someone decided to intervene when it came time for reprints.

And here's the panel after clean-up and corrections. I've corrected the earlier corrections by cloning Dave's actual letter forms from other portions of the page and flown them in to assist here. It can take some finessing to make this look natural. In this case, the "TA" in "TAVERN" are the original letters, but I had to shift them over to the right for them to look natural aside their replacement neighbors. Having now had almost two month's worth of experience, I see more I would do here now, but I think it's clear this is a vast improvement on the prior printing.

I'd spent several weeks working on, and talking about, a Photoshop action script that would automate the page production, sharpening and adjusting the scans so that the files were prepped with a minimum of by-hand adjustment.

And it worked! It worked beautifully with the High Society negatives I had access to, it worked with the original art we had, it worked with the book scans I tested it with.

And then, I tried it with newsprint. Disaster.

Oh, it worked alright. Every tiny little line was retained, every nuance that had made it through to the page.

Unfortunately, the sharpening also brought up every little blemish, every piece of noise in the original

newsprint.

And newsprint can be really noisy. 

For one, the detail we were picking up from the pamphlet printings is at least partially due to very light coverage of ink from the printing plate, which also meant sometimes uneven coverage of large areas of black. In the scan, this manifests itself as little freckles of white inside of the black, little freckles that expand and grow in prominence as they are sharpened and contrast-adjusted.

The second type of noise is pulpy bits of unbleached paper which, once contrast-adjusted, look an

awful lot like intentionally-made lines or dots.

Fortunately these noise issues are solvable problems, even if the solutions available take time.

Above: An unprocessed newsprint scan from issue 13, with blacks typical of that issue.
Below: A panel from the same issue after processing and clean-up. 

Basically, I tried two different solutions to this. One was to add a step to the script that functioned as a noise reduction stage. I used the Photoshop function "surface blur," with carefully selected radius and threshold settings, which did an excellent job of wiping out most of the noise. Unfortunately, on pages with very fine white on black lines, this also had a tendency to wipe out some detail. Once again, this meant sharpening these detail areas by hand.

This solution was less viable on pages that had any amount of airbrushing/spatter, as the small spatter areas are interpreted as noise by the algorithm.

On these pages, I used a second solution, sweeping the solid black areas with the "burn" tool set on mid-tones. I also experimented with masking those areas and then running noise reduction on the remainder of the page.

There are several other possible solution involving more involved masking and the like, but, as I've said many times, these are the kinds of solutions that are practical when you're dealing with 5 pages, or 50 pages, or even 500 pages. Not the kinds of things you want to be doing with 6,000 pages.

That being said, the automated approach, with the aid of some surface blur, is present on the finished pages. Early in the morning on deadline day, plowing through the remainder of the pages, I let the computer handle 5 pages all by itself. Take a look at 114-118 if you'd like to see.

Those five pages are a good example of the most economic "restoration" possible, an example of scanning the best material available, and using the automation routines I've developed, pointing the computer at them and saying "go."

How then do these pages differ from the rest?

Hows about we start with panel borders?

Let's hear from Dave Sim, circa 1981, describing friend and fellow cartoonist Gene Day.

...he has always crushed me with his sense of black. Where to put it. How much. He leaves the cap off so it will get thicker. There is nothing even remotely resembling a washed-out black on a Gene Day original. And he is so clean! Look at the corners of a Gene Day panel. They're corners! Square -- the corners come to a point. Look at the corners of a Dave Sim panel (on second thought don't look at the corners of a Dave Sim panel).

Throughout most of the four signatures we prepared for this book, I took the (perhaps unusual) step of touching up the panel borders. Specifically, stray lines that wandered out of bounds, over-extensions, any kind of marks that violate the gutter space of the pages.

Like most aesthetic effects that rely on an accumulation of visual detail, it's a little difficult to

demonstrate this on screen. I think page 111 is a good example of this, so take out your original printing and the new one to compare.

A closeup of a single panel of the original, with no correction. Taken in isolation this doesn't look particularly bad, but a page full of ragged unintentional overlapping areas adds visual "noise" to what started as a very dynamic and purposeful design.

 Below is my fairly conservative "cleaning" of this panel. I've addressed most of the overlapping panel border elements and the brush overlap/edge tremor, but left in-panel examples of the same alone. This is how it now appears in print.

Below is an example of how far you could take this work if you wanted. I've eliminated in-panel examples of the same kinds of tangents from all of the spaces obviously intended to be clean-edged. This is the level of cleanup that I would do on my own artwork, although it's possible this veers into "face lift" territory when working on historical material.

To me, anyway, the effect is one of cleaning up the page, of unifying it a bit more than it was originally. The eye is free to focus on movements within the panel, and the tangents created by the stray lines aren't there anymore to distract focus. (Once again, consult the printed pages, as it's much more dramatic there than on-screen, even in blow-up).

I consulted with Dave about this before I did it, but I didn't go crazy with it this time around, mainly because I wanted to get his (and your) reactions to this level of adjustment before I took it any further.

Notice these are not perfectly squared-off, mechanically 90 degree corners. That I'm afraid is not how human beings produce corners, not how ink reacts in real-life, and thus not how I'm squaring them off. There's still hand-made, still human, still how Dave drew them, just not out of boundaries of the panels.

For me there is an aesthetic gain from this that's well worth the 5-10 minute outlay of time per page that it costs. This is also a problem that virtually disappears, say, 100 pages into Church and State I, as Dave has become a more disciplined craftsman, and increasingly relies on Letratape for borders.

So, we've had some significant success here. As promised, no moire patterns, no detail left behind, incredibly crisp fine-line information with no breakup whatsoever, except when that breakup is in the source material. It's a huge jump from where things were at just a few short months ago.

So what can we do better for next time?

First off, we didn't get as much cleanup as we'd like. There is still "schmutz" in a few places, surviving

artifacts of newsprint that we either didn't notice or thought were actual information rather than noise. Some of it is in between panels, some is in white space, some is little white flecks in areas of pure black.

There are three basic solutions to this. One, do clean-up after the files had been grayscale converted. (For whatever reason, doing the grayscale conversion seems to bring up a little more previously invisible noise, even though there was already a threshold adjustment layer in place). Second, use a nice "gutter sweep" technique Lou Copeland wrote me about, to get any garbage outside of the panels in one pass. Lastly, not prep pages all night with no chance for revision! (I'm afraid the last one was the real culprit here).

The second real thing to talk about is the issue of tone shrinkage.

Take a look at page 18. This, like a handful of pages here, was sourced from original art, in this case, a scan appropriated from the Heritage Auction website.

Before:

After:

You'll see that in the bottom middle panel, Cerebus is missing some of his tone above his sword. (fortunately, it happens to look light a lighting effect as it is now...)

This is because the Letratone used to make Cerebus' halftone, and various other background effects, was made up of a very thin acetate with emulsion on top, and a sticky backing. Over time this acetate shrinks while the rest of the page remains the same size, causing the tone to either pull away from its original position, or if it was particularly well-burnished in the first place, to rip in half.

I don't know the whys and wheres of tone shrinkage, why certain pages I've seen have none while others have so much. It seems somewhat a function of age, possibly relating to tone size, somewhat a function of some other factor (humidity? light exposure?) that I don't know about. But on a practical level, what it means is, essentially, a lot of work for someone.

Issue 6 (pages 119-140) in the new printing is a good demonstration of both the benefits and drawbacks of working straight from aged original art without time for adjustment.

In these pages I've adjusted only what I considered the most distracting examples of the tone

shrinkage, which I've done by cloning adjacent areas of tone in Photoshop and flying it over to fill in the gaps. This is trickier than you might think, as any slight deviation in dot angle is very, very visible in print, even if it's invisible on screen.

A close-up of a panel on page 134, prior to tone cloning. 

After tone cloning. 

Both Lou Copeland and George Gatsis suggested all-out replacing tone in several instances as well, in places where the tone is damaged by newsprint or even not cut well in the first place. Once again, these are problems that will largely disappear as we get past the first two books.

But they are significant problems, in the financial sense, if nothing else. And Dave is looking to you all, the Cerebus patrons/life support system, for guidance. How much is it worth? If it cost $40 extra per page sourced from original artwork to have all of the tone as clean as clean can be, is that a good use of your money? How about $100, in the most extreme cases?

I think that somewhere between $20 and $60 a page (1/2 to 1 + 1/2 hour of work) is probably the most likely range for pages like these, with so much tone shrinkage, but, as with anything, it's likely I'll get faster, and it's equally likely we'll butt up against some previously undiscovered problem that'll slow things down again. As an example, take a look at the top right corner of page 137. The mezzotint tone on the statue has split and fractured. How do you patch that, short of drawing out all of the tone and dropping a whole new virtual sheet of mezzotint on top of it? The pattern's too irregular to just drag an adjacent area. Any other ideas? The finer dot mezzotint is a little more forgiving, but the coarser stuff needs a careful eye for adjustment. (In the weeks since I've written this, I have solved this one, but it's still a slow solution, unfortunately.)

The second thing to do differently, and the thing that most definitely will be improved in future printings, is adjusting the exposure of the pages evenly so that Cerebus' tone is consistent, no matter the exposure of the source. I won't go into any detail about this now, because I'll be writing about it in the coming weeks. That being said, it's an area I've done a ton of thinking about, and have made a lot of progress since preparing these replacement signatures.

The last problem is an issue that's for the most part out of my control. That's the issue of black density/ink coverage.

On my copies of the book, the ink coverage is on the low end of normal, the blacks a little closer to the "gray" portion of the spectrum, especially compared to the extremely bright white of the paper. The good news on this is it won't be the case for next time. Lebonfon has instituted a series of changes in their press and their ink formula, which they're calling "the Aardvark Initiative," that have done wonders for the ink density. So when we get to High Society, this should no longer be an issue. (If indeed it is here, these things being largely a matter of taste and personal frame of reference.)

*********************************************************************************

For me, the past two months of work have been a blizzard of learning, and a transformation in attitude. Many of the recommendations I made initially, for instance, working almost exclusively from negatives and printed material and ignoring the original art unless there was compelling reason to retreat to it, were primarily based on economics and efficiency. What's the most expedient way to do this? What's the most reasonable way to do this?

But what I've gradually come around to is the realization that there's more to be had. More detail, more refinement. That the closer you get to the source, the clearer things become.

It's a strange thing to look at these pages now. It's only been two months, but I've gained a tremendous amount of experience since then. It's hard to see these pages now for how they are, but rather for how they can be in the future.

And that's really the excitement of this entire enterprise to me. To take an amazing piece of art, and help preserve it, and present it in the best way possible. I'd like to thank all of you for that opportunity.

Comments! Suggestions! Questions! Please! Below!